Cattle livestock trading is scaling with exponential growth and man-aging this scaled production and procurement has become a hugechallenge for the industry. With a vision to solve these challengeswith emerging technologies and state-of-the-art techniques fromcomputer vision, we tried to develop a large possible dataset thatis labeled and accurate so that our research community can buildtheir novel models or finetune existing ones without the hassle ofdata collection. We developed our cow images dataset of 17,899images with vitals (sex, color, breed, feed, age, teeth, height, weight,price, size) which can be used for both classification and regression.We are also contributing baseline models demonstrating how thisdataset can be used for regression and classification. These baselinemodels consist of multi input-output network(1 input - 2 outputs;3 outputs; 4 outputs) to classify and regress cattle livestock vitalsamong them (1 input - 2 outputs) have the best accuracy of 75% and67% respectively for age and breed with the minimum loss. Estimat-ing or predicting cattle livestock vitals is an open research area andour cow images dataset and baseline models are going to play a vitalrole towards further research opportunity. Repository link for thispaper is attached here https://github.com/bhuiyanmobasshir94/CID
Dataset creation repository - Cow-weight-and-Breed-Prediction